SARMAE: Masked Autoencoder for SAR Representation Learning

2025-12-18

AI Summary

To address the limitations in fine-grained semantic representation learning for SAR imagery—stemming from data scarcity and inherent speckle noise—the paper proposes a Noise-Aware Masked Autoencoder (NAE) framework. Methodologically, it integrates self-supervised pretraining, contrastive learning, and noise-robust reconstruction. Key contributions include: (1) introducing SAR-1M, the first high-quality, million-scale SAR dataset; (2) designing a Speckle-Aware Representation Enhancement (SARE) module that explicitly models speckle noise distribution; and (3) proposing Semantic Anchored Regularization Constraints (SARC) to enforce cross-modal feature alignment under optical–SAR paired supervision. Evaluated on classification, detection, and segmentation tasks, NAE achieves state-of-the-art performance across all three, significantly advancing fine-grained semantic understanding of SAR imagery and enhancing the generalizability of all-weather remote sensing representations.

Technology Category

• Computer Vision: Representation Learning for Vision
• Machine Learning: Representation Learning
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Web query analysis, representation and understanding
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs

Abstract

Synthetic Aperture Radar (SAR) imagery plays a critical role in all-weather, day-and-night remote sensing applications. However, existing SAR-oriented deep learning is constrained by data scarcity, while the physically grounded speckle noise in SAR imagery further hampers fine-grained semantic representation learning. To address these challenges, we propose SARMAE, a Noise-Aware Masked Autoencoder for self-supervised SAR representation learning. Specifically, we construct SAR-1M, the first million-scale SAR dataset, with additional paired optical images, to enable large-scale pre-training. Building upon this, we design Speckle-Aware Representation Enhancement (SARE), which injects SAR-specific speckle noise into masked autoencoders to facilitate noise-aware and robust representation learning. Furthermore, we introduce Semantic Anchor Representation Constraint (SARC), which leverages paired optical priors to align SAR features and ensure semantic consistency. Extensive experiments across multiple SAR datasets demonstrate that SARMAE achieves state-of-the-art performance on classification, detection, and segmentation tasks. Code and models will be available at https://github.com/MiliLab/SARMAE.

Problem

• Addresses data scarcity in SAR deep learning
• Mitigates speckle noise for robust representation learning
• Ensures semantic consistency using paired optical priors

Innovation

• Constructs million-scale SAR dataset with optical images
• Injects speckle noise into autoencoders for robustness
• Uses optical priors to align SAR features semantically